Hacker News Reader

Comment by jsheard | Hacker News Reader

AIUI the spec being leaked ironically makes things worse, because for an unofficial implementation to be legally kosher it would have to be clean-room reverse engineered anyway, and since the official spec is out there the integrity of such an effort would be called into doubt. You'd somehow have to prove you didn't look at it, ever, or at least be trusted enough for people to take your word for it.

(I'm not a lawyer, please correct me if I'm wrong)

View on HN

Reading a standards spec to understand what the device you paid for does?

Straight to jail!

Pirating the entire internet to train your AI?

That's fair use.

Can we just train an AI with the spec and then vibe code an implementation?

Depends on if you can fund a defense all the way to the Supreme Court.

Just get said AI to write it yourself for my own hardware.. come get me HDMI law nerds!

I have thought about writing a python web framework were instead of writing a function that handles a request, you write a docstring and the an AI JIT generates your handling code. Could we not just prompt-engineer a solution for the missing bits in the driver for the HDMI2 stuff and have it be lazily generated via a model parameter and an API key? And then, in about 10 years, we could just do it locally once the models become runnable on commodity hardware. What a future too look forward to.

This is not a web framework, but it's pretty close to what you've imagined: https://github.com/JirkaKlimes/jit-implementation

I hope someone can do this in such a manner as to engineer the collision of the legal titans. Either way, we win on some ground.

IP vs AI, round two, Fight!!!

> IP vs AI, round two, Fight!!!

I want to hear an EPIC RAP BATTLE OF HISTORY version of this.

ChatGPT VERSus HDMI Forum BEGIN!

Changes are the leaked spec might already be in the popular LLMs' training data, so you could probably skip to step 2 without having to do the (potentially problematic) training yourself.

Are you ridiculing the concept of imaginary property?

It does make sense. If you are on the money receiving side.

On the other side: do you pay license fees to your parents, your teachers, ... everybody you ever learnt from? No? Why not? Didn't everybody learn by copying first?

What about imitation? What does freedom of art and science even mean? You call it parody. I call it theft.

See. You need the contradionary concept of imaginary property. Otherwise, how do we get rich quick? Live performance, consultation, teaching? Nah, those are for loosers... Rent seeking it is.

They're wrong, there's nothing stopping you implementing anything you like, you just can't use the HDMI brand without complying with their rules.

This sounds too easy to be true.

Does the "brand" include the physical shape of the connector?

Could I make hardware with a "NotHDMI" port that "happens" to be mechanically compatible with HDMI plugs, has the exact same pinout, etc etc?

Even then: In the OP case the hardware is already there, it's only about the driver. So wouldn't a driver for hardware that very clearly identifies the port as "HDMI" run into the same problem, even if the driver itself never mentions the term?

I mean, have you seen "TF Card" slots?

I’ve been waiting for the “TF card” version of HDMI for a while.

You reminded me of the flipper zero video game module[0] with it's "video out port" which "transmits a video signal in DVI-D format to an external TV, monitor, or projector".

They are not quite the size of Valve though, and can expect people to figure out what that that port is.

0: https://docs.flipper.net/zero/video-game-module

They just used the well-known PicoDVI implementation that exists for the Raspberry Pi Pico:

> https://picockpit.com/raspberry-pi/raspberry-pi-pico-video-o... (scroll down to "DVI")

> https://github.com/Wren6991/PicoDVI

No, the connectors wouldn't be regulated, you're not violating any IP by buying them and there's no prohibition on any of the manufacturers selling them to unlicensed companies. At worst you can assert a patent against the design but there's no specific patent for that design, there are patents for some aspects of the design/implementation but they're hold by the manufacturers of the connectors themselves.

There have been many examples in the past of consumer electronics companies selling things that are electrically and logically compatible with HDMI, but they just have to avoid using the word HDMI.

Probably one thing that the HDMI forum is holding over AMD/Valve is that there's an API to manage some of the functions of the HDMI driver. They could infer that this API is a part of the closed standards of HDMI Forum. But 90% of the threat is about certification and branding I am sure.

> Could I make hardware with a "NotHDMI" port that "happens" to be mechanically compatible with HDMI plugs, has the exact same pinout, etc etc?

Yes there are a bunch of products that do exactly this, sometimes with the same pin out and used for video output to HDMI compatible screens (internal HDMI mods for consoles are an example), other ones that use it for completely different purposes like controller ports (the Bliss-Box adapters and MiSTeR SNAC controllers). They just can’t use the HDMI name. In fact a few of those HDMI console mods started with HDMI in their product names and changes them for exactly that reason (e.g. DCHDMI now called DCDigital for the Sega Dreamcast).

> you just can't use the HDMI brand

Hmmm, okay so maybe HDNI: High-Definition Not Incumbered.

HAHI, which stands for HAHI Ain't an HDMI Implementation.

That's why we have to train LLMs to infringe patents and implement them. That's fair use by their own logic.

The game is getting sued by the HDMI forum. It doesn't matter how "clean" your implementation was. They're just going to sue you _anyways_.

Innocence until proven guilty should mean the burden of proof is on the prosecutor to prove you actually looked at it right? Although that isn't necessarily how it works in the real world. IANAL.

But I also don't understand how they would enforce that you can't use a leaked spec. If there are patents involved that would hinder an open source implementation regardless of if it was clean room or not. I don't think copyright would apply, because the implementation is not the same as the spec. And trademark would only apply if you used hdmi branding materials (so just say something like "this driver provides compatibility with an interface that has been hostile to open source that starts with h and ends with i"), and if you use a leaked spec, you didn't sign any contracts saying you can't implement it.

> Innocence until proven guilty should mean the burden of proof is on the prosecutor to prove you actually looked at it right?

It wouldn't be criminal, just civil, and civil trials have much lower standard for the burden of proof. It's just preponderance of evidence (more likely than not), instead of beyond all reasonable doubt.

The level of proof required is lower, but AFAIU, the burden of proof is on the prosecutor, not the defendent.

IIUC, the problem is a bit tautological. Regardless of legality of reverse engineering itself, HDMI is a trademark which you obviously cannot use without being licensed. Using HDMI connector itself is probably a grey-ish area: while you can buy the connectors without agreeing to any licenses and forwarding compliance on vendor, it would still be hard to argue that you had no idea it was a HDMI connector. If you are using the HDMI connector, but are not sending anything else but DVI over it, it should be fine-ish.

The real problem starts when you want to actually support HDMI 2.0 and 2.1 on top. Arguing that you have licenced for 2.0 and then tacked a clean-room implementation of 2.1 on top gets essentially impossible.

For stuff like connectors, this gets worked around by using terminology like “compatible with HDMI” all the time. You are explicitly permitted to reference your competitor’s products, including potential compatibility, by trademark law. I suspect the risk here is mostly contractural - AMD likely signed agreements with the HDMI forum a long time ago that restrict their disclosure of details from the specification.

Im shocked i had to scroll so far to find a real hard stop blocker mentioned.

Valve has no reason to care about using the HDMI trademark. Consumers dont care if it says HDMI 2.1 or HMDI 2.1 Compatible.

The connector isnt trademarked and neither is compatibility.

The oss nature of isnt one either as valve could just release a compiled binary instead of open sourcing it.

The 'get sued for copying the leak' argument implies someone would actually fancy going toe to toe with valves legal team which so far have rekt the eu, activision, riot games, microsoft, etc. in court.

Proving beyond doubt that valve or their devs accessed the leaks would be hard. Especially if valve were clever from the get go, and lets face it, they probably were. Theyre easily one of the leanest, most profitable, and savviest software companies around.

The connector itself shouldn't be an issue, because it doesn't fall under IP. The shape of the connector is entirely functional, so there's no creative work involved, so it would fall under patent law. However, the connector itself is unlikely to be innovative enough to be patentable, so it's not protected by patent law either.

Using HDMI connectors is totally fine. You just can't label it as "has HDMI port", as "HDMI" is a trademark.

Is that true? There is obviously some creative work in connector design - optimizing for looks, robustness to damage, dirt, easy of use, reliability technically, etc.

HDMI's gate is certification and the ability to then use their marketing brand.

This is absolutely not a technical issue. You can implement the 2.1 spec if you want, you just can't say it's 2.1.

If Valve wanted they could happily get it to work and let people figure out that it works, they just can't use that title in their marketing.

IIUC the issue is not them being unable to implement 2.1 at all, but rather provide specifically open source implementation. They probably could provide a binary blob.

That's probably how NVidia did it.

But there's very little software involved in HDMI, it's mostly hardware and a control API.

Yes, this is exactly what the "proprietary" driver is doing

I've seen HDMI devices for sale on AliExpress that list their port as "HDMI-compatible" or just "HD" to avoid that certification requirement.

AFAIK clean-room reverse engineering is sufficient but not always necessary for such an implementation to be allowed, but it does make the fair use argument a bit more difficult. (and of course the DMCA criminalizes any reverse engineering of 'technical safeguards' regardless of how you do it)

The spec is open to them and this isn't an IP issue, it's a branding issue.

Clean room RE isn't legally required. It just makes a stronger defense against claims of infringement.

They don't really have to worry about patent infringement, the biggest issue is that they can implement anything they want, they just can't call it HDMI 2.1 without certification.

That's confusing for the consumer but technically viable.

HDMI exists to write standards, to certify them and to enforce the brand integrity. Patents are a different issue and would be handled separately.

(I am an engineer who spent half his career dealing with this stuff at a technical, legal and commercial level).

> they just can't call it HDMI 2.1 without certification

The problem is more that they can't use the HDMI trademark at all, not just for the HDMI 2.1 on Linux implementation. That makes it a non-starter for AMD or Valve, but in theory should not stop an individual who doesn't care about marketing anything as being HDMI-compatible.

Clean room reverse engineering produces specification when you don't have it. When you have specification, you don't have to reverse engineer it.

>You'd somehow have to prove you didn't look at it, ever, or at least be trusted enough for people to take your word for it.

How could one prove a negative? It's supposed to be innocent until proven guilty, isn't it? They'd have to prove that you've looked at the spec files.

> innocent until proven guilty

In the US at least, for criminal cases, the burden of proof for guilt is "beyond a reasonable doubt". For civil cases they are much more lenient and use the "reasonable person" standard.

> You'd somehow have to prove you didn't look at it

You can't prove something that didn't happen, unless you were monitored your whole life or at least from the moment the item came into being. It's an unreasonable level of proof.

Probably now or the very near future you could have an LLM that's provably trained on dataset where the leaked spec isn't included in the dataset and have it perform the reverse engineering work.

As someone who’s excited to see this happen eventually, it’s not happening anytime soon. Combinatorial optimization techniques are far better suited for this and methods created 50 years ago run laps around LLMs

Reply on news.ycombinator.com